4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/init.h>
57 #include <linux/capability.h>
58 #include <linux/file.h>
59 #include <linux/string.h>
60 #include <linux/seq_file.h>
61 #include <linux/namei.h>
62 #include <linux/mnt_namespace.h>
64 #include <linux/rcupdate.h>
65 #include <linux/kallsyms.h>
66 #include <linux/module.h>
67 #include <linux/mount.h>
68 #include <linux/security.h>
69 #include <linux/ptrace.h>
70 #include <linux/seccomp.h>
71 #include <linux/cpuset.h>
72 #include <linux/audit.h>
73 #include <linux/poll.h>
74 #include <linux/nsproxy.h>
75 #include <linux/oom.h>
79 * Implementing inode permission operations in /proc is almost
80 * certainly an error. Permission checks need to happen during
81 * each system call not at open time. The reason is that most of
82 * what we wish to check for permissions in /proc varies at runtime.
84 * The classic example of a problem is opening file descriptors
85 * in /proc for a task before it execs a suid executable.
89 /* Worst case buffer size needed for holding an integer. */
90 #define PROC_NUMBUF 13
96 const struct inode_operations
*iop
;
97 const struct file_operations
*fop
;
101 #define NOD(NAME, MODE, IOP, FOP, OP) { \
103 .len = sizeof(NAME) - 1, \
110 #define DIR(NAME, MODE, OTYPE) \
111 NOD(NAME, (S_IFDIR|(MODE)), \
112 &proc_##OTYPE##_inode_operations, &proc_##OTYPE##_operations, \
114 #define LNK(NAME, OTYPE) \
115 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
116 &proc_pid_link_inode_operations, NULL, \
117 { .proc_get_link = &proc_##OTYPE##_link } )
118 #define REG(NAME, MODE, OTYPE) \
119 NOD(NAME, (S_IFREG|(MODE)), NULL, \
120 &proc_##OTYPE##_operations, {})
121 #define INF(NAME, MODE, OTYPE) \
122 NOD(NAME, (S_IFREG|(MODE)), \
123 NULL, &proc_info_file_operations, \
124 { .proc_read = &proc_##OTYPE } )
127 EXPORT_SYMBOL(maps_protect
);
129 static struct fs_struct
*get_fs_struct(struct task_struct
*task
)
131 struct fs_struct
*fs
;
135 atomic_inc(&fs
->count
);
140 static int get_nr_threads(struct task_struct
*tsk
)
142 /* Must be called with the rcu_read_lock held */
146 if (lock_task_sighand(tsk
, &flags
)) {
147 count
= atomic_read(&tsk
->signal
->count
);
148 unlock_task_sighand(tsk
, &flags
);
153 static int proc_cwd_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
155 struct task_struct
*task
= get_proc_task(inode
);
156 struct fs_struct
*fs
= NULL
;
157 int result
= -ENOENT
;
160 fs
= get_fs_struct(task
);
161 put_task_struct(task
);
164 read_lock(&fs
->lock
);
165 *mnt
= mntget(fs
->pwdmnt
);
166 *dentry
= dget(fs
->pwd
);
167 read_unlock(&fs
->lock
);
174 static int proc_root_link(struct inode
*inode
, struct dentry
**dentry
, struct vfsmount
**mnt
)
176 struct task_struct
*task
= get_proc_task(inode
);
177 struct fs_struct
*fs
= NULL
;
178 int result
= -ENOENT
;
181 fs
= get_fs_struct(task
);
182 put_task_struct(task
);
185 read_lock(&fs
->lock
);
186 *mnt
= mntget(fs
->rootmnt
);
187 *dentry
= dget(fs
->root
);
188 read_unlock(&fs
->lock
);
195 #define MAY_PTRACE(task) \
196 (task == current || \
197 (task->parent == current && \
198 (task->ptrace & PT_PTRACED) && \
199 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
200 security_ptrace(current,task) == 0))
202 static int proc_pid_environ(struct task_struct
*task
, char * buffer
)
205 struct mm_struct
*mm
= get_task_mm(task
);
207 unsigned int len
= mm
->env_end
- mm
->env_start
;
210 res
= access_process_vm(task
, mm
->env_start
, buffer
, len
, 0);
211 if (!ptrace_may_attach(task
))
218 static int proc_pid_cmdline(struct task_struct
*task
, char * buffer
)
222 struct mm_struct
*mm
= get_task_mm(task
);
226 goto out_mm
; /* Shh! No looking before we're done */
228 len
= mm
->arg_end
- mm
->arg_start
;
233 res
= access_process_vm(task
, mm
->arg_start
, buffer
, len
, 0);
235 // If the nul at the end of args has been overwritten, then
236 // assume application is using setproctitle(3).
237 if (res
> 0 && buffer
[res
-1] != '\0' && len
< PAGE_SIZE
) {
238 len
= strnlen(buffer
, res
);
242 len
= mm
->env_end
- mm
->env_start
;
243 if (len
> PAGE_SIZE
- res
)
244 len
= PAGE_SIZE
- res
;
245 res
+= access_process_vm(task
, mm
->env_start
, buffer
+res
, len
, 0);
246 res
= strnlen(buffer
, res
);
255 static int proc_pid_auxv(struct task_struct
*task
, char *buffer
)
258 struct mm_struct
*mm
= get_task_mm(task
);
260 unsigned int nwords
= 0;
263 while (mm
->saved_auxv
[nwords
- 2] != 0); /* AT_NULL */
264 res
= nwords
* sizeof(mm
->saved_auxv
[0]);
267 memcpy(buffer
, mm
->saved_auxv
, res
);
274 #ifdef CONFIG_KALLSYMS
276 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
277 * Returns the resolved symbol. If that fails, simply return the address.
279 static int proc_pid_wchan(struct task_struct
*task
, char *buffer
)
282 char symname
[KSYM_NAME_LEN
+1];
284 wchan
= get_wchan(task
);
286 if (lookup_symbol_name(wchan
, symname
) < 0)
287 return sprintf(buffer
, "%lu", wchan
);
289 return sprintf(buffer
, "%s", symname
);
291 #endif /* CONFIG_KALLSYMS */
293 #ifdef CONFIG_SCHEDSTATS
295 * Provides /proc/PID/schedstat
297 static int proc_pid_schedstat(struct task_struct
*task
, char *buffer
)
299 return sprintf(buffer
, "%llu %llu %lu\n",
300 task
->sched_info
.cpu_time
,
301 task
->sched_info
.run_delay
,
302 task
->sched_info
.pcnt
);
306 /* The badness from the OOM killer */
307 unsigned long badness(struct task_struct
*p
, unsigned long uptime
);
308 static int proc_oom_score(struct task_struct
*task
, char *buffer
)
310 unsigned long points
;
311 struct timespec uptime
;
313 do_posix_clock_monotonic_gettime(&uptime
);
314 read_lock(&tasklist_lock
);
315 points
= badness(task
, uptime
.tv_sec
);
316 read_unlock(&tasklist_lock
);
317 return sprintf(buffer
, "%lu\n", points
);
320 /************************************************************************/
321 /* Here the fs part begins */
322 /************************************************************************/
324 /* permission checks */
325 static int proc_fd_access_allowed(struct inode
*inode
)
327 struct task_struct
*task
;
329 /* Allow access to a task's file descriptors if it is us or we
330 * may use ptrace attach to the process and find out that
333 task
= get_proc_task(inode
);
335 allowed
= ptrace_may_attach(task
);
336 put_task_struct(task
);
341 static int proc_setattr(struct dentry
*dentry
, struct iattr
*attr
)
344 struct inode
*inode
= dentry
->d_inode
;
346 if (attr
->ia_valid
& ATTR_MODE
)
349 error
= inode_change_ok(inode
, attr
);
351 error
= inode_setattr(inode
, attr
);
355 static const struct inode_operations proc_def_inode_operations
= {
356 .setattr
= proc_setattr
,
359 extern struct seq_operations mounts_op
;
365 static int mounts_open(struct inode
*inode
, struct file
*file
)
367 struct task_struct
*task
= get_proc_task(inode
);
368 struct mnt_namespace
*ns
= NULL
;
369 struct proc_mounts
*p
;
375 ns
= task
->nsproxy
->mnt_ns
;
380 put_task_struct(task
);
385 p
= kmalloc(sizeof(struct proc_mounts
), GFP_KERNEL
);
387 file
->private_data
= &p
->m
;
388 ret
= seq_open(file
, &mounts_op
);
391 p
->event
= ns
->event
;
401 static int mounts_release(struct inode
*inode
, struct file
*file
)
403 struct seq_file
*m
= file
->private_data
;
404 struct mnt_namespace
*ns
= m
->private;
406 return seq_release(inode
, file
);
409 static unsigned mounts_poll(struct file
*file
, poll_table
*wait
)
411 struct proc_mounts
*p
= file
->private_data
;
412 struct mnt_namespace
*ns
= p
->m
.private;
415 poll_wait(file
, &ns
->poll
, wait
);
417 spin_lock(&vfsmount_lock
);
418 if (p
->event
!= ns
->event
) {
419 p
->event
= ns
->event
;
422 spin_unlock(&vfsmount_lock
);
427 static const struct file_operations proc_mounts_operations
= {
431 .release
= mounts_release
,
435 extern struct seq_operations mountstats_op
;
436 static int mountstats_open(struct inode
*inode
, struct file
*file
)
438 int ret
= seq_open(file
, &mountstats_op
);
441 struct seq_file
*m
= file
->private_data
;
442 struct mnt_namespace
*mnt_ns
= NULL
;
443 struct task_struct
*task
= get_proc_task(inode
);
448 mnt_ns
= task
->nsproxy
->mnt_ns
;
452 put_task_struct(task
);
458 seq_release(inode
, file
);
465 static const struct file_operations proc_mountstats_operations
= {
466 .open
= mountstats_open
,
469 .release
= mounts_release
,
472 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
474 static ssize_t
proc_info_read(struct file
* file
, char __user
* buf
,
475 size_t count
, loff_t
*ppos
)
477 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
480 struct task_struct
*task
= get_proc_task(inode
);
486 if (count
> PROC_BLOCK_SIZE
)
487 count
= PROC_BLOCK_SIZE
;
490 if (!(page
= __get_free_page(GFP_KERNEL
)))
493 length
= PROC_I(inode
)->op
.proc_read(task
, (char*)page
);
496 length
= simple_read_from_buffer(buf
, count
, ppos
, (char *)page
, length
);
499 put_task_struct(task
);
504 static const struct file_operations proc_info_file_operations
= {
505 .read
= proc_info_read
,
508 static int mem_open(struct inode
* inode
, struct file
* file
)
510 file
->private_data
= (void*)((long)current
->self_exec_id
);
514 static ssize_t
mem_read(struct file
* file
, char __user
* buf
,
515 size_t count
, loff_t
*ppos
)
517 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
519 unsigned long src
= *ppos
;
521 struct mm_struct
*mm
;
526 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
530 page
= (char *)__get_free_page(GFP_USER
);
536 mm
= get_task_mm(task
);
542 if (file
->private_data
!= (void*)((long)current
->self_exec_id
))
548 int this_len
, retval
;
550 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
551 retval
= access_process_vm(task
, src
, page
, this_len
, 0);
552 if (!retval
|| !MAY_PTRACE(task
) || !ptrace_may_attach(task
)) {
558 if (copy_to_user(buf
, page
, retval
)) {
573 free_page((unsigned long) page
);
575 put_task_struct(task
);
580 #define mem_write NULL
583 /* This is a security hazard */
584 static ssize_t
mem_write(struct file
* file
, const char __user
*buf
,
585 size_t count
, loff_t
*ppos
)
589 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
590 unsigned long dst
= *ppos
;
596 if (!MAY_PTRACE(task
) || !ptrace_may_attach(task
))
600 page
= (char *)__get_free_page(GFP_USER
);
606 int this_len
, retval
;
608 this_len
= (count
> PAGE_SIZE
) ? PAGE_SIZE
: count
;
609 if (copy_from_user(page
, buf
, this_len
)) {
613 retval
= access_process_vm(task
, dst
, page
, this_len
, 1);
625 free_page((unsigned long) page
);
627 put_task_struct(task
);
633 static loff_t
mem_lseek(struct file
* file
, loff_t offset
, int orig
)
637 file
->f_pos
= offset
;
640 file
->f_pos
+= offset
;
645 force_successful_syscall_return();
649 static const struct file_operations proc_mem_operations
= {
656 static ssize_t
oom_adjust_read(struct file
*file
, char __user
*buf
,
657 size_t count
, loff_t
*ppos
)
659 struct task_struct
*task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
660 char buffer
[PROC_NUMBUF
];
666 oom_adjust
= task
->oomkilladj
;
667 put_task_struct(task
);
669 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", oom_adjust
);
671 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
674 static ssize_t
oom_adjust_write(struct file
*file
, const char __user
*buf
,
675 size_t count
, loff_t
*ppos
)
677 struct task_struct
*task
;
678 char buffer
[PROC_NUMBUF
], *end
;
681 memset(buffer
, 0, sizeof(buffer
));
682 if (count
> sizeof(buffer
) - 1)
683 count
= sizeof(buffer
) - 1;
684 if (copy_from_user(buffer
, buf
, count
))
686 oom_adjust
= simple_strtol(buffer
, &end
, 0);
687 if ((oom_adjust
< OOM_ADJUST_MIN
|| oom_adjust
> OOM_ADJUST_MAX
) &&
688 oom_adjust
!= OOM_DISABLE
)
692 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
695 if (oom_adjust
< task
->oomkilladj
&& !capable(CAP_SYS_RESOURCE
)) {
696 put_task_struct(task
);
699 task
->oomkilladj
= oom_adjust
;
700 put_task_struct(task
);
701 if (end
- buffer
== 0)
706 static const struct file_operations proc_oom_adjust_operations
= {
707 .read
= oom_adjust_read
,
708 .write
= oom_adjust_write
,
712 static ssize_t
clear_refs_write(struct file
*file
, const char __user
*buf
,
713 size_t count
, loff_t
*ppos
)
715 struct task_struct
*task
;
716 char buffer
[PROC_NUMBUF
], *end
;
717 struct mm_struct
*mm
;
719 memset(buffer
, 0, sizeof(buffer
));
720 if (count
> sizeof(buffer
) - 1)
721 count
= sizeof(buffer
) - 1;
722 if (copy_from_user(buffer
, buf
, count
))
724 if (!simple_strtol(buffer
, &end
, 0))
728 task
= get_proc_task(file
->f_path
.dentry
->d_inode
);
731 mm
= get_task_mm(task
);
736 put_task_struct(task
);
737 if (end
- buffer
== 0)
742 static struct file_operations proc_clear_refs_operations
= {
743 .write
= clear_refs_write
,
747 #ifdef CONFIG_AUDITSYSCALL
749 static ssize_t
proc_loginuid_read(struct file
* file
, char __user
* buf
,
750 size_t count
, loff_t
*ppos
)
752 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
753 struct task_struct
*task
= get_proc_task(inode
);
755 char tmpbuf
[TMPBUFLEN
];
759 length
= scnprintf(tmpbuf
, TMPBUFLEN
, "%u",
760 audit_get_loginuid(task
->audit_context
));
761 put_task_struct(task
);
762 return simple_read_from_buffer(buf
, count
, ppos
, tmpbuf
, length
);
765 static ssize_t
proc_loginuid_write(struct file
* file
, const char __user
* buf
,
766 size_t count
, loff_t
*ppos
)
768 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
773 if (!capable(CAP_AUDIT_CONTROL
))
776 if (current
!= pid_task(proc_pid(inode
), PIDTYPE_PID
))
779 if (count
>= PAGE_SIZE
)
780 count
= PAGE_SIZE
- 1;
783 /* No partial writes. */
786 page
= (char*)__get_free_page(GFP_USER
);
790 if (copy_from_user(page
, buf
, count
))
794 loginuid
= simple_strtoul(page
, &tmp
, 10);
800 length
= audit_set_loginuid(current
, loginuid
);
801 if (likely(length
== 0))
805 free_page((unsigned long) page
);
809 static const struct file_operations proc_loginuid_operations
= {
810 .read
= proc_loginuid_read
,
811 .write
= proc_loginuid_write
,
815 #ifdef CONFIG_SECCOMP
816 static ssize_t
seccomp_read(struct file
*file
, char __user
*buf
,
817 size_t count
, loff_t
*ppos
)
819 struct task_struct
*tsk
= get_proc_task(file
->f_dentry
->d_inode
);
825 /* no need to print the trailing zero, so use only len */
826 len
= sprintf(__buf
, "%u\n", tsk
->seccomp
.mode
);
827 put_task_struct(tsk
);
829 return simple_read_from_buffer(buf
, count
, ppos
, __buf
, len
);
832 static ssize_t
seccomp_write(struct file
*file
, const char __user
*buf
,
833 size_t count
, loff_t
*ppos
)
835 struct task_struct
*tsk
= get_proc_task(file
->f_dentry
->d_inode
);
836 char __buf
[20], *end
;
837 unsigned int seccomp_mode
;
844 /* can set it only once to be even more secure */
846 if (unlikely(tsk
->seccomp
.mode
))
850 memset(__buf
, 0, sizeof(__buf
));
851 count
= min(count
, sizeof(__buf
) - 1);
852 if (copy_from_user(__buf
, buf
, count
))
855 seccomp_mode
= simple_strtoul(__buf
, &end
, 0);
859 if (seccomp_mode
&& seccomp_mode
<= NR_SECCOMP_MODES
) {
860 tsk
->seccomp
.mode
= seccomp_mode
;
861 set_tsk_thread_flag(tsk
, TIF_SECCOMP
);
865 if (unlikely(!(end
- __buf
)))
867 result
= end
- __buf
;
869 put_task_struct(tsk
);
874 static const struct file_operations proc_seccomp_operations
= {
875 .read
= seccomp_read
,
876 .write
= seccomp_write
,
878 #endif /* CONFIG_SECCOMP */
880 #ifdef CONFIG_FAULT_INJECTION
881 static ssize_t
proc_fault_inject_read(struct file
* file
, char __user
* buf
,
882 size_t count
, loff_t
*ppos
)
884 struct task_struct
*task
= get_proc_task(file
->f_dentry
->d_inode
);
885 char buffer
[PROC_NUMBUF
];
891 make_it_fail
= task
->make_it_fail
;
892 put_task_struct(task
);
894 len
= snprintf(buffer
, sizeof(buffer
), "%i\n", make_it_fail
);
896 return simple_read_from_buffer(buf
, count
, ppos
, buffer
, len
);
899 static ssize_t
proc_fault_inject_write(struct file
* file
,
900 const char __user
* buf
, size_t count
, loff_t
*ppos
)
902 struct task_struct
*task
;
903 char buffer
[PROC_NUMBUF
], *end
;
906 if (!capable(CAP_SYS_RESOURCE
))
908 memset(buffer
, 0, sizeof(buffer
));
909 if (count
> sizeof(buffer
) - 1)
910 count
= sizeof(buffer
) - 1;
911 if (copy_from_user(buffer
, buf
, count
))
913 make_it_fail
= simple_strtol(buffer
, &end
, 0);
916 task
= get_proc_task(file
->f_dentry
->d_inode
);
919 task
->make_it_fail
= make_it_fail
;
920 put_task_struct(task
);
921 if (end
- buffer
== 0)
926 static const struct file_operations proc_fault_inject_operations
= {
927 .read
= proc_fault_inject_read
,
928 .write
= proc_fault_inject_write
,
932 #ifdef CONFIG_SCHED_DEBUG
934 * Print out various scheduling related per-task fields:
936 static int sched_show(struct seq_file
*m
, void *v
)
938 struct inode
*inode
= m
->private;
939 struct task_struct
*p
;
943 p
= get_proc_task(inode
);
946 proc_sched_show_task(p
, m
);
954 sched_write(struct file
*file
, const char __user
*buf
,
955 size_t count
, loff_t
*offset
)
957 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
958 struct task_struct
*p
;
962 p
= get_proc_task(inode
);
965 proc_sched_set_task(p
);
972 static int sched_open(struct inode
*inode
, struct file
*filp
)
976 ret
= single_open(filp
, sched_show
, NULL
);
978 struct seq_file
*m
= filp
->private_data
;
985 static const struct file_operations proc_pid_sched_operations
= {
988 .write
= sched_write
,
990 .release
= seq_release
,
995 static void *proc_pid_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
997 struct inode
*inode
= dentry
->d_inode
;
1000 /* We don't need a base pointer in the /proc filesystem */
1003 /* Are we allowed to snoop on the tasks file descriptors? */
1004 if (!proc_fd_access_allowed(inode
))
1007 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &nd
->dentry
, &nd
->mnt
);
1008 nd
->last_type
= LAST_BIND
;
1010 return ERR_PTR(error
);
1013 static int do_proc_readlink(struct dentry
*dentry
, struct vfsmount
*mnt
,
1014 char __user
*buffer
, int buflen
)
1016 struct inode
* inode
;
1017 char *tmp
= (char*)__get_free_page(GFP_KERNEL
), *path
;
1023 inode
= dentry
->d_inode
;
1024 path
= d_path(dentry
, mnt
, tmp
, PAGE_SIZE
);
1025 len
= PTR_ERR(path
);
1028 len
= tmp
+ PAGE_SIZE
- 1 - path
;
1032 if (copy_to_user(buffer
, path
, len
))
1035 free_page((unsigned long)tmp
);
1039 static int proc_pid_readlink(struct dentry
* dentry
, char __user
* buffer
, int buflen
)
1041 int error
= -EACCES
;
1042 struct inode
*inode
= dentry
->d_inode
;
1044 struct vfsmount
*mnt
= NULL
;
1046 /* Are we allowed to snoop on the tasks file descriptors? */
1047 if (!proc_fd_access_allowed(inode
))
1050 error
= PROC_I(inode
)->op
.proc_get_link(inode
, &de
, &mnt
);
1054 error
= do_proc_readlink(de
, mnt
, buffer
, buflen
);
1061 static const struct inode_operations proc_pid_link_inode_operations
= {
1062 .readlink
= proc_pid_readlink
,
1063 .follow_link
= proc_pid_follow_link
,
1064 .setattr
= proc_setattr
,
1068 /* building an inode */
1070 static int task_dumpable(struct task_struct
*task
)
1073 struct mm_struct
*mm
;
1078 dumpable
= mm
->dumpable
;
1086 static struct inode
*proc_pid_make_inode(struct super_block
* sb
, struct task_struct
*task
)
1088 struct inode
* inode
;
1089 struct proc_inode
*ei
;
1091 /* We need a new inode */
1093 inode
= new_inode(sb
);
1099 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1100 inode
->i_op
= &proc_def_inode_operations
;
1103 * grab the reference to task.
1105 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1111 if (task_dumpable(task
)) {
1112 inode
->i_uid
= task
->euid
;
1113 inode
->i_gid
= task
->egid
;
1115 security_task_to_inode(task
, inode
);
1125 static int pid_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1127 struct inode
*inode
= dentry
->d_inode
;
1128 struct task_struct
*task
;
1129 generic_fillattr(inode
, stat
);
1134 task
= pid_task(proc_pid(inode
), PIDTYPE_PID
);
1136 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1137 task_dumpable(task
)) {
1138 stat
->uid
= task
->euid
;
1139 stat
->gid
= task
->egid
;
1149 * Exceptional case: normally we are not allowed to unhash a busy
1150 * directory. In this case, however, we can do it - no aliasing problems
1151 * due to the way we treat inodes.
1153 * Rewrite the inode's ownerships here because the owning task may have
1154 * performed a setuid(), etc.
1156 * Before the /proc/pid/status file was created the only way to read
1157 * the effective uid of a /process was to stat /proc/pid. Reading
1158 * /proc/pid/status is slow enough that procps and other packages
1159 * kept stating /proc/pid. To keep the rules in /proc simple I have
1160 * made this apply to all per process world readable and executable
1163 static int pid_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1165 struct inode
*inode
= dentry
->d_inode
;
1166 struct task_struct
*task
= get_proc_task(inode
);
1168 if ((inode
->i_mode
== (S_IFDIR
|S_IRUGO
|S_IXUGO
)) ||
1169 task_dumpable(task
)) {
1170 inode
->i_uid
= task
->euid
;
1171 inode
->i_gid
= task
->egid
;
1176 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1177 security_task_to_inode(task
, inode
);
1178 put_task_struct(task
);
1185 static int pid_delete_dentry(struct dentry
* dentry
)
1187 /* Is the task we represent dead?
1188 * If so, then don't put the dentry on the lru list,
1189 * kill it immediately.
1191 return !proc_pid(dentry
->d_inode
)->tasks
[PIDTYPE_PID
].first
;
1194 static struct dentry_operations pid_dentry_operations
=
1196 .d_revalidate
= pid_revalidate
,
1197 .d_delete
= pid_delete_dentry
,
1202 typedef struct dentry
*instantiate_t(struct inode
*, struct dentry
*,
1203 struct task_struct
*, const void *);
1206 * Fill a directory entry.
1208 * If possible create the dcache entry and derive our inode number and
1209 * file type from dcache entry.
1211 * Since all of the proc inode numbers are dynamically generated, the inode
1212 * numbers do not exist until the inode is cache. This means creating the
1213 * the dcache entry in readdir is necessary to keep the inode numbers
1214 * reported by readdir in sync with the inode numbers reported
1217 static int proc_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
1218 char *name
, int len
,
1219 instantiate_t instantiate
, struct task_struct
*task
, const void *ptr
)
1221 struct dentry
*child
, *dir
= filp
->f_path
.dentry
;
1222 struct inode
*inode
;
1225 unsigned type
= DT_UNKNOWN
;
1229 qname
.hash
= full_name_hash(name
, len
);
1231 child
= d_lookup(dir
, &qname
);
1234 new = d_alloc(dir
, &qname
);
1236 child
= instantiate(dir
->d_inode
, new, task
, ptr
);
1243 if (!child
|| IS_ERR(child
) || !child
->d_inode
)
1244 goto end_instantiate
;
1245 inode
= child
->d_inode
;
1248 type
= inode
->i_mode
>> 12;
1253 ino
= find_inode_number(dir
, &qname
);
1256 return filldir(dirent
, name
, len
, filp
->f_pos
, ino
, type
);
1259 static unsigned name_to_int(struct dentry
*dentry
)
1261 const char *name
= dentry
->d_name
.name
;
1262 int len
= dentry
->d_name
.len
;
1265 if (len
> 1 && *name
== '0')
1268 unsigned c
= *name
++ - '0';
1271 if (n
>= (~0U-9)/10)
1281 #define PROC_FDINFO_MAX 64
1283 static int proc_fd_info(struct inode
*inode
, struct dentry
**dentry
,
1284 struct vfsmount
**mnt
, char *info
)
1286 struct task_struct
*task
= get_proc_task(inode
);
1287 struct files_struct
*files
= NULL
;
1289 int fd
= proc_fd(inode
);
1292 files
= get_files_struct(task
);
1293 put_task_struct(task
);
1297 * We are not taking a ref to the file structure, so we must
1300 spin_lock(&files
->file_lock
);
1301 file
= fcheck_files(files
, fd
);
1304 *mnt
= mntget(file
->f_path
.mnt
);
1306 *dentry
= dget(file
->f_path
.dentry
);
1308 snprintf(info
, PROC_FDINFO_MAX
,
1311 (long long) file
->f_pos
,
1313 spin_unlock(&files
->file_lock
);
1314 put_files_struct(files
);
1317 spin_unlock(&files
->file_lock
);
1318 put_files_struct(files
);
1323 static int proc_fd_link(struct inode
*inode
, struct dentry
**dentry
,
1324 struct vfsmount
**mnt
)
1326 return proc_fd_info(inode
, dentry
, mnt
, NULL
);
1329 static int tid_fd_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1331 struct inode
*inode
= dentry
->d_inode
;
1332 struct task_struct
*task
= get_proc_task(inode
);
1333 int fd
= proc_fd(inode
);
1334 struct files_struct
*files
;
1337 files
= get_files_struct(task
);
1340 if (fcheck_files(files
, fd
)) {
1342 put_files_struct(files
);
1343 if (task_dumpable(task
)) {
1344 inode
->i_uid
= task
->euid
;
1345 inode
->i_gid
= task
->egid
;
1350 inode
->i_mode
&= ~(S_ISUID
| S_ISGID
);
1351 security_task_to_inode(task
, inode
);
1352 put_task_struct(task
);
1356 put_files_struct(files
);
1358 put_task_struct(task
);
1364 static struct dentry_operations tid_fd_dentry_operations
=
1366 .d_revalidate
= tid_fd_revalidate
,
1367 .d_delete
= pid_delete_dentry
,
1370 static struct dentry
*proc_fd_instantiate(struct inode
*dir
,
1371 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1373 unsigned fd
= *(const unsigned *)ptr
;
1375 struct files_struct
*files
;
1376 struct inode
*inode
;
1377 struct proc_inode
*ei
;
1378 struct dentry
*error
= ERR_PTR(-ENOENT
);
1380 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1385 files
= get_files_struct(task
);
1388 inode
->i_mode
= S_IFLNK
;
1391 * We are not taking a ref to the file structure, so we must
1394 spin_lock(&files
->file_lock
);
1395 file
= fcheck_files(files
, fd
);
1398 if (file
->f_mode
& 1)
1399 inode
->i_mode
|= S_IRUSR
| S_IXUSR
;
1400 if (file
->f_mode
& 2)
1401 inode
->i_mode
|= S_IWUSR
| S_IXUSR
;
1402 spin_unlock(&files
->file_lock
);
1403 put_files_struct(files
);
1405 inode
->i_op
= &proc_pid_link_inode_operations
;
1407 ei
->op
.proc_get_link
= proc_fd_link
;
1408 dentry
->d_op
= &tid_fd_dentry_operations
;
1409 d_add(dentry
, inode
);
1410 /* Close the race of the process dying before we return the dentry */
1411 if (tid_fd_revalidate(dentry
, NULL
))
1417 spin_unlock(&files
->file_lock
);
1418 put_files_struct(files
);
1424 static struct dentry
*proc_lookupfd_common(struct inode
*dir
,
1425 struct dentry
*dentry
,
1426 instantiate_t instantiate
)
1428 struct task_struct
*task
= get_proc_task(dir
);
1429 unsigned fd
= name_to_int(dentry
);
1430 struct dentry
*result
= ERR_PTR(-ENOENT
);
1437 result
= instantiate(dir
, dentry
, task
, &fd
);
1439 put_task_struct(task
);
1444 static int proc_readfd_common(struct file
* filp
, void * dirent
,
1445 filldir_t filldir
, instantiate_t instantiate
)
1447 struct dentry
*dentry
= filp
->f_path
.dentry
;
1448 struct inode
*inode
= dentry
->d_inode
;
1449 struct task_struct
*p
= get_proc_task(inode
);
1450 unsigned int fd
, tid
, ino
;
1452 struct files_struct
* files
;
1453 struct fdtable
*fdt
;
1464 if (filldir(dirent
, ".", 1, 0, inode
->i_ino
, DT_DIR
) < 0)
1468 ino
= parent_ino(dentry
);
1469 if (filldir(dirent
, "..", 2, 1, ino
, DT_DIR
) < 0)
1473 files
= get_files_struct(p
);
1477 fdt
= files_fdtable(files
);
1478 for (fd
= filp
->f_pos
-2;
1480 fd
++, filp
->f_pos
++) {
1481 char name
[PROC_NUMBUF
];
1484 if (!fcheck_files(files
, fd
))
1488 len
= snprintf(name
, sizeof(name
), "%d", fd
);
1489 if (proc_fill_cache(filp
, dirent
, filldir
,
1490 name
, len
, instantiate
,
1498 put_files_struct(files
);
1506 static struct dentry
*proc_lookupfd(struct inode
*dir
, struct dentry
*dentry
,
1507 struct nameidata
*nd
)
1509 return proc_lookupfd_common(dir
, dentry
, proc_fd_instantiate
);
1512 static int proc_readfd(struct file
*filp
, void *dirent
, filldir_t filldir
)
1514 return proc_readfd_common(filp
, dirent
, filldir
, proc_fd_instantiate
);
1517 static ssize_t
proc_fdinfo_read(struct file
*file
, char __user
*buf
,
1518 size_t len
, loff_t
*ppos
)
1520 char tmp
[PROC_FDINFO_MAX
];
1521 int err
= proc_fd_info(file
->f_path
.dentry
->d_inode
, NULL
, NULL
, tmp
);
1523 err
= simple_read_from_buffer(buf
, len
, ppos
, tmp
, strlen(tmp
));
1527 static const struct file_operations proc_fdinfo_file_operations
= {
1528 .open
= nonseekable_open
,
1529 .read
= proc_fdinfo_read
,
1532 static const struct file_operations proc_fd_operations
= {
1533 .read
= generic_read_dir
,
1534 .readdir
= proc_readfd
,
1538 * /proc/pid/fd needs a special permission handler so that a process can still
1539 * access /proc/self/fd after it has executed a setuid().
1541 static int proc_fd_permission(struct inode
*inode
, int mask
,
1542 struct nameidata
*nd
)
1546 rv
= generic_permission(inode
, mask
, NULL
);
1549 if (task_pid(current
) == proc_pid(inode
))
1555 * proc directories can do almost nothing..
1557 static const struct inode_operations proc_fd_inode_operations
= {
1558 .lookup
= proc_lookupfd
,
1559 .permission
= proc_fd_permission
,
1560 .setattr
= proc_setattr
,
1563 static struct dentry
*proc_fdinfo_instantiate(struct inode
*dir
,
1564 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1566 unsigned fd
= *(unsigned *)ptr
;
1567 struct inode
*inode
;
1568 struct proc_inode
*ei
;
1569 struct dentry
*error
= ERR_PTR(-ENOENT
);
1571 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1576 inode
->i_mode
= S_IFREG
| S_IRUSR
;
1577 inode
->i_fop
= &proc_fdinfo_file_operations
;
1578 dentry
->d_op
= &tid_fd_dentry_operations
;
1579 d_add(dentry
, inode
);
1580 /* Close the race of the process dying before we return the dentry */
1581 if (tid_fd_revalidate(dentry
, NULL
))
1588 static struct dentry
*proc_lookupfdinfo(struct inode
*dir
,
1589 struct dentry
*dentry
,
1590 struct nameidata
*nd
)
1592 return proc_lookupfd_common(dir
, dentry
, proc_fdinfo_instantiate
);
1595 static int proc_readfdinfo(struct file
*filp
, void *dirent
, filldir_t filldir
)
1597 return proc_readfd_common(filp
, dirent
, filldir
,
1598 proc_fdinfo_instantiate
);
1601 static const struct file_operations proc_fdinfo_operations
= {
1602 .read
= generic_read_dir
,
1603 .readdir
= proc_readfdinfo
,
1607 * proc directories can do almost nothing..
1609 static const struct inode_operations proc_fdinfo_inode_operations
= {
1610 .lookup
= proc_lookupfdinfo
,
1611 .setattr
= proc_setattr
,
1615 static struct dentry
*proc_pident_instantiate(struct inode
*dir
,
1616 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1618 const struct pid_entry
*p
= ptr
;
1619 struct inode
*inode
;
1620 struct proc_inode
*ei
;
1621 struct dentry
*error
= ERR_PTR(-EINVAL
);
1623 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
1628 inode
->i_mode
= p
->mode
;
1629 if (S_ISDIR(inode
->i_mode
))
1630 inode
->i_nlink
= 2; /* Use getattr to fix if necessary */
1632 inode
->i_op
= p
->iop
;
1634 inode
->i_fop
= p
->fop
;
1636 dentry
->d_op
= &pid_dentry_operations
;
1637 d_add(dentry
, inode
);
1638 /* Close the race of the process dying before we return the dentry */
1639 if (pid_revalidate(dentry
, NULL
))
1645 static struct dentry
*proc_pident_lookup(struct inode
*dir
,
1646 struct dentry
*dentry
,
1647 const struct pid_entry
*ents
,
1650 struct inode
*inode
;
1651 struct dentry
*error
;
1652 struct task_struct
*task
= get_proc_task(dir
);
1653 const struct pid_entry
*p
, *last
;
1655 error
= ERR_PTR(-ENOENT
);
1662 * Yes, it does not scale. And it should not. Don't add
1663 * new entries into /proc/<tgid>/ without very good reasons.
1665 last
= &ents
[nents
- 1];
1666 for (p
= ents
; p
<= last
; p
++) {
1667 if (p
->len
!= dentry
->d_name
.len
)
1669 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
1675 error
= proc_pident_instantiate(dir
, dentry
, task
, p
);
1677 put_task_struct(task
);
1682 static int proc_pident_fill_cache(struct file
*filp
, void *dirent
,
1683 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
1685 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
1686 proc_pident_instantiate
, task
, p
);
1689 static int proc_pident_readdir(struct file
*filp
,
1690 void *dirent
, filldir_t filldir
,
1691 const struct pid_entry
*ents
, unsigned int nents
)
1695 struct dentry
*dentry
= filp
->f_path
.dentry
;
1696 struct inode
*inode
= dentry
->d_inode
;
1697 struct task_struct
*task
= get_proc_task(inode
);
1698 const struct pid_entry
*p
, *last
;
1712 if (filldir(dirent
, ".", 1, i
, ino
, DT_DIR
) < 0)
1718 ino
= parent_ino(dentry
);
1719 if (filldir(dirent
, "..", 2, i
, ino
, DT_DIR
) < 0)
1731 last
= &ents
[nents
- 1];
1733 if (proc_pident_fill_cache(filp
, dirent
, filldir
, task
, p
) < 0)
1742 put_task_struct(task
);
1747 #ifdef CONFIG_SECURITY
1748 static ssize_t
proc_pid_attr_read(struct file
* file
, char __user
* buf
,
1749 size_t count
, loff_t
*ppos
)
1751 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1754 struct task_struct
*task
= get_proc_task(inode
);
1759 length
= security_getprocattr(task
,
1760 (char*)file
->f_path
.dentry
->d_name
.name
,
1762 put_task_struct(task
);
1764 length
= simple_read_from_buffer(buf
, count
, ppos
, p
, length
);
1769 static ssize_t
proc_pid_attr_write(struct file
* file
, const char __user
* buf
,
1770 size_t count
, loff_t
*ppos
)
1772 struct inode
* inode
= file
->f_path
.dentry
->d_inode
;
1775 struct task_struct
*task
= get_proc_task(inode
);
1780 if (count
> PAGE_SIZE
)
1783 /* No partial writes. */
1789 page
= (char*)__get_free_page(GFP_USER
);
1794 if (copy_from_user(page
, buf
, count
))
1797 length
= security_setprocattr(task
,
1798 (char*)file
->f_path
.dentry
->d_name
.name
,
1799 (void*)page
, count
);
1801 free_page((unsigned long) page
);
1803 put_task_struct(task
);
1808 static const struct file_operations proc_pid_attr_operations
= {
1809 .read
= proc_pid_attr_read
,
1810 .write
= proc_pid_attr_write
,
1813 static const struct pid_entry attr_dir_stuff
[] = {
1814 REG("current", S_IRUGO
|S_IWUGO
, pid_attr
),
1815 REG("prev", S_IRUGO
, pid_attr
),
1816 REG("exec", S_IRUGO
|S_IWUGO
, pid_attr
),
1817 REG("fscreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1818 REG("keycreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1819 REG("sockcreate", S_IRUGO
|S_IWUGO
, pid_attr
),
1822 static int proc_attr_dir_readdir(struct file
* filp
,
1823 void * dirent
, filldir_t filldir
)
1825 return proc_pident_readdir(filp
,dirent
,filldir
,
1826 attr_dir_stuff
,ARRAY_SIZE(attr_dir_stuff
));
1829 static const struct file_operations proc_attr_dir_operations
= {
1830 .read
= generic_read_dir
,
1831 .readdir
= proc_attr_dir_readdir
,
1834 static struct dentry
*proc_attr_dir_lookup(struct inode
*dir
,
1835 struct dentry
*dentry
, struct nameidata
*nd
)
1837 return proc_pident_lookup(dir
, dentry
,
1838 attr_dir_stuff
, ARRAY_SIZE(attr_dir_stuff
));
1841 static const struct inode_operations proc_attr_dir_inode_operations
= {
1842 .lookup
= proc_attr_dir_lookup
,
1843 .getattr
= pid_getattr
,
1844 .setattr
= proc_setattr
,
1852 static int proc_self_readlink(struct dentry
*dentry
, char __user
*buffer
,
1855 char tmp
[PROC_NUMBUF
];
1856 sprintf(tmp
, "%d", current
->tgid
);
1857 return vfs_readlink(dentry
,buffer
,buflen
,tmp
);
1860 static void *proc_self_follow_link(struct dentry
*dentry
, struct nameidata
*nd
)
1862 char tmp
[PROC_NUMBUF
];
1863 sprintf(tmp
, "%d", current
->tgid
);
1864 return ERR_PTR(vfs_follow_link(nd
,tmp
));
1867 static const struct inode_operations proc_self_inode_operations
= {
1868 .readlink
= proc_self_readlink
,
1869 .follow_link
= proc_self_follow_link
,
1875 * These are the directory entries in the root directory of /proc
1876 * that properly belong to the /proc filesystem, as they describe
1877 * describe something that is process related.
1879 static const struct pid_entry proc_base_stuff
[] = {
1880 NOD("self", S_IFLNK
|S_IRWXUGO
,
1881 &proc_self_inode_operations
, NULL
, {}),
1885 * Exceptional case: normally we are not allowed to unhash a busy
1886 * directory. In this case, however, we can do it - no aliasing problems
1887 * due to the way we treat inodes.
1889 static int proc_base_revalidate(struct dentry
*dentry
, struct nameidata
*nd
)
1891 struct inode
*inode
= dentry
->d_inode
;
1892 struct task_struct
*task
= get_proc_task(inode
);
1894 put_task_struct(task
);
1901 static struct dentry_operations proc_base_dentry_operations
=
1903 .d_revalidate
= proc_base_revalidate
,
1904 .d_delete
= pid_delete_dentry
,
1907 static struct dentry
*proc_base_instantiate(struct inode
*dir
,
1908 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
1910 const struct pid_entry
*p
= ptr
;
1911 struct inode
*inode
;
1912 struct proc_inode
*ei
;
1913 struct dentry
*error
= ERR_PTR(-EINVAL
);
1915 /* Allocate the inode */
1916 error
= ERR_PTR(-ENOMEM
);
1917 inode
= new_inode(dir
->i_sb
);
1921 /* Initialize the inode */
1923 inode
->i_mtime
= inode
->i_atime
= inode
->i_ctime
= CURRENT_TIME
;
1926 * grab the reference to the task.
1928 ei
->pid
= get_task_pid(task
, PIDTYPE_PID
);
1934 inode
->i_mode
= p
->mode
;
1935 if (S_ISDIR(inode
->i_mode
))
1937 if (S_ISLNK(inode
->i_mode
))
1940 inode
->i_op
= p
->iop
;
1942 inode
->i_fop
= p
->fop
;
1944 dentry
->d_op
= &proc_base_dentry_operations
;
1945 d_add(dentry
, inode
);
1954 static struct dentry
*proc_base_lookup(struct inode
*dir
, struct dentry
*dentry
)
1956 struct dentry
*error
;
1957 struct task_struct
*task
= get_proc_task(dir
);
1958 const struct pid_entry
*p
, *last
;
1960 error
= ERR_PTR(-ENOENT
);
1965 /* Lookup the directory entry */
1966 last
= &proc_base_stuff
[ARRAY_SIZE(proc_base_stuff
) - 1];
1967 for (p
= proc_base_stuff
; p
<= last
; p
++) {
1968 if (p
->len
!= dentry
->d_name
.len
)
1970 if (!memcmp(dentry
->d_name
.name
, p
->name
, p
->len
))
1976 error
= proc_base_instantiate(dir
, dentry
, task
, p
);
1979 put_task_struct(task
);
1984 static int proc_base_fill_cache(struct file
*filp
, void *dirent
,
1985 filldir_t filldir
, struct task_struct
*task
, const struct pid_entry
*p
)
1987 return proc_fill_cache(filp
, dirent
, filldir
, p
->name
, p
->len
,
1988 proc_base_instantiate
, task
, p
);
1991 #ifdef CONFIG_TASK_IO_ACCOUNTING
1992 static int proc_pid_io_accounting(struct task_struct
*task
, char *buffer
)
1994 return sprintf(buffer
,
1995 #ifdef CONFIG_TASK_XACCT
2001 "read_bytes: %llu\n"
2002 "write_bytes: %llu\n"
2003 "cancelled_write_bytes: %llu\n",
2004 #ifdef CONFIG_TASK_XACCT
2005 (unsigned long long)task
->rchar
,
2006 (unsigned long long)task
->wchar
,
2007 (unsigned long long)task
->syscr
,
2008 (unsigned long long)task
->syscw
,
2010 (unsigned long long)task
->ioac
.read_bytes
,
2011 (unsigned long long)task
->ioac
.write_bytes
,
2012 (unsigned long long)task
->ioac
.cancelled_write_bytes
);
2019 static const struct file_operations proc_task_operations
;
2020 static const struct inode_operations proc_task_inode_operations
;
2022 static const struct pid_entry tgid_base_stuff
[] = {
2023 DIR("task", S_IRUGO
|S_IXUGO
, task
),
2024 DIR("fd", S_IRUSR
|S_IXUSR
, fd
),
2025 DIR("fdinfo", S_IRUSR
|S_IXUSR
, fdinfo
),
2026 INF("environ", S_IRUSR
, pid_environ
),
2027 INF("auxv", S_IRUSR
, pid_auxv
),
2028 INF("status", S_IRUGO
, pid_status
),
2029 #ifdef CONFIG_SCHED_DEBUG
2030 REG("sched", S_IRUGO
|S_IWUSR
, pid_sched
),
2032 INF("cmdline", S_IRUGO
, pid_cmdline
),
2033 INF("stat", S_IRUGO
, tgid_stat
),
2034 INF("statm", S_IRUGO
, pid_statm
),
2035 REG("maps", S_IRUGO
, maps
),
2037 REG("numa_maps", S_IRUGO
, numa_maps
),
2039 REG("mem", S_IRUSR
|S_IWUSR
, mem
),
2040 #ifdef CONFIG_SECCOMP
2041 REG("seccomp", S_IRUSR
|S_IWUSR
, seccomp
),
2046 REG("mounts", S_IRUGO
, mounts
),
2047 REG("mountstats", S_IRUSR
, mountstats
),
2049 REG("clear_refs", S_IWUSR
, clear_refs
),
2050 REG("smaps", S_IRUGO
, smaps
),
2052 #ifdef CONFIG_SECURITY
2053 DIR("attr", S_IRUGO
|S_IXUGO
, attr_dir
),
2055 #ifdef CONFIG_KALLSYMS
2056 INF("wchan", S_IRUGO
, pid_wchan
),
2058 #ifdef CONFIG_SCHEDSTATS
2059 INF("schedstat", S_IRUGO
, pid_schedstat
),
2061 #ifdef CONFIG_CPUSETS
2062 REG("cpuset", S_IRUGO
, cpuset
),
2064 INF("oom_score", S_IRUGO
, oom_score
),
2065 REG("oom_adj", S_IRUGO
|S_IWUSR
, oom_adjust
),
2066 #ifdef CONFIG_AUDITSYSCALL
2067 REG("loginuid", S_IWUSR
|S_IRUGO
, loginuid
),
2069 #ifdef CONFIG_FAULT_INJECTION
2070 REG("make-it-fail", S_IRUGO
|S_IWUSR
, fault_inject
),
2072 #ifdef CONFIG_TASK_IO_ACCOUNTING
2073 INF("io", S_IRUGO
, pid_io_accounting
),
2077 static int proc_tgid_base_readdir(struct file
* filp
,
2078 void * dirent
, filldir_t filldir
)
2080 return proc_pident_readdir(filp
,dirent
,filldir
,
2081 tgid_base_stuff
,ARRAY_SIZE(tgid_base_stuff
));
2084 static const struct file_operations proc_tgid_base_operations
= {
2085 .read
= generic_read_dir
,
2086 .readdir
= proc_tgid_base_readdir
,
2089 static struct dentry
*proc_tgid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2090 return proc_pident_lookup(dir
, dentry
,
2091 tgid_base_stuff
, ARRAY_SIZE(tgid_base_stuff
));
2094 static const struct inode_operations proc_tgid_base_inode_operations
= {
2095 .lookup
= proc_tgid_base_lookup
,
2096 .getattr
= pid_getattr
,
2097 .setattr
= proc_setattr
,
2101 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2103 * @task: task that should be flushed.
2105 * Looks in the dcache for
2107 * /proc/@tgid/task/@pid
2108 * if either directory is present flushes it and all of it'ts children
2111 * It is safe and reasonable to cache /proc entries for a task until
2112 * that task exits. After that they just clog up the dcache with
2113 * useless entries, possibly causing useful dcache entries to be
2114 * flushed instead. This routine is proved to flush those useless
2115 * dcache entries at process exit time.
2117 * NOTE: This routine is just an optimization so it does not guarantee
2118 * that no dcache entries will exist at process exit time it
2119 * just makes it very unlikely that any will persist.
2121 void proc_flush_task(struct task_struct
*task
)
2123 struct dentry
*dentry
, *leader
, *dir
;
2124 char buf
[PROC_NUMBUF
];
2128 name
.len
= snprintf(buf
, sizeof(buf
), "%d", task
->pid
);
2129 dentry
= d_hash_and_lookup(proc_mnt
->mnt_root
, &name
);
2131 shrink_dcache_parent(dentry
);
2136 if (thread_group_leader(task
))
2140 name
.len
= snprintf(buf
, sizeof(buf
), "%d", task
->tgid
);
2141 leader
= d_hash_and_lookup(proc_mnt
->mnt_root
, &name
);
2146 name
.len
= strlen(name
.name
);
2147 dir
= d_hash_and_lookup(leader
, &name
);
2149 goto out_put_leader
;
2152 name
.len
= snprintf(buf
, sizeof(buf
), "%d", task
->pid
);
2153 dentry
= d_hash_and_lookup(dir
, &name
);
2155 shrink_dcache_parent(dentry
);
2167 static struct dentry
*proc_pid_instantiate(struct inode
*dir
,
2168 struct dentry
* dentry
,
2169 struct task_struct
*task
, const void *ptr
)
2171 struct dentry
*error
= ERR_PTR(-ENOENT
);
2172 struct inode
*inode
;
2174 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2178 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2179 inode
->i_op
= &proc_tgid_base_inode_operations
;
2180 inode
->i_fop
= &proc_tgid_base_operations
;
2181 inode
->i_flags
|=S_IMMUTABLE
;
2183 #ifdef CONFIG_SECURITY
2184 inode
->i_nlink
+= 1;
2187 dentry
->d_op
= &pid_dentry_operations
;
2189 d_add(dentry
, inode
);
2190 /* Close the race of the process dying before we return the dentry */
2191 if (pid_revalidate(dentry
, NULL
))
2197 struct dentry
*proc_pid_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2199 struct dentry
*result
= ERR_PTR(-ENOENT
);
2200 struct task_struct
*task
;
2203 result
= proc_base_lookup(dir
, dentry
);
2204 if (!IS_ERR(result
) || PTR_ERR(result
) != -ENOENT
)
2207 tgid
= name_to_int(dentry
);
2212 task
= find_task_by_pid(tgid
);
2214 get_task_struct(task
);
2219 result
= proc_pid_instantiate(dir
, dentry
, task
, NULL
);
2220 put_task_struct(task
);
2226 * Find the first task with tgid >= tgid
2229 static struct task_struct
*next_tgid(unsigned int tgid
)
2231 struct task_struct
*task
;
2237 pid
= find_ge_pid(tgid
);
2240 task
= pid_task(pid
, PIDTYPE_PID
);
2241 /* What we to know is if the pid we have find is the
2242 * pid of a thread_group_leader. Testing for task
2243 * being a thread_group_leader is the obvious thing
2244 * todo but there is a window when it fails, due to
2245 * the pid transfer logic in de_thread.
2247 * So we perform the straight forward test of seeing
2248 * if the pid we have found is the pid of a thread
2249 * group leader, and don't worry if the task we have
2250 * found doesn't happen to be a thread group leader.
2251 * As we don't care in the case of readdir.
2253 if (!task
|| !has_group_leader_pid(task
))
2255 get_task_struct(task
);
2261 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2263 static int proc_pid_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
2264 struct task_struct
*task
, int tgid
)
2266 char name
[PROC_NUMBUF
];
2267 int len
= snprintf(name
, sizeof(name
), "%d", tgid
);
2268 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
2269 proc_pid_instantiate
, task
, NULL
);
2272 /* for the /proc/ directory itself, after non-process stuff has been done */
2273 int proc_pid_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2275 unsigned int nr
= filp
->f_pos
- FIRST_PROCESS_ENTRY
;
2276 struct task_struct
*reaper
= get_proc_task(filp
->f_path
.dentry
->d_inode
);
2277 struct task_struct
*task
;
2283 for (; nr
< ARRAY_SIZE(proc_base_stuff
); filp
->f_pos
++, nr
++) {
2284 const struct pid_entry
*p
= &proc_base_stuff
[nr
];
2285 if (proc_base_fill_cache(filp
, dirent
, filldir
, reaper
, p
) < 0)
2289 tgid
= filp
->f_pos
- TGID_OFFSET
;
2290 for (task
= next_tgid(tgid
);
2292 put_task_struct(task
), task
= next_tgid(tgid
+ 1)) {
2294 filp
->f_pos
= tgid
+ TGID_OFFSET
;
2295 if (proc_pid_fill_cache(filp
, dirent
, filldir
, task
, tgid
) < 0) {
2296 put_task_struct(task
);
2300 filp
->f_pos
= PID_MAX_LIMIT
+ TGID_OFFSET
;
2302 put_task_struct(reaper
);
2310 static const struct pid_entry tid_base_stuff
[] = {
2311 DIR("fd", S_IRUSR
|S_IXUSR
, fd
),
2312 DIR("fdinfo", S_IRUSR
|S_IXUSR
, fdinfo
),
2313 INF("environ", S_IRUSR
, pid_environ
),
2314 INF("auxv", S_IRUSR
, pid_auxv
),
2315 INF("status", S_IRUGO
, pid_status
),
2316 #ifdef CONFIG_SCHED_DEBUG
2317 REG("sched", S_IRUGO
|S_IWUSR
, pid_sched
),
2319 INF("cmdline", S_IRUGO
, pid_cmdline
),
2320 INF("stat", S_IRUGO
, tid_stat
),
2321 INF("statm", S_IRUGO
, pid_statm
),
2322 REG("maps", S_IRUGO
, maps
),
2324 REG("numa_maps", S_IRUGO
, numa_maps
),
2326 REG("mem", S_IRUSR
|S_IWUSR
, mem
),
2327 #ifdef CONFIG_SECCOMP
2328 REG("seccomp", S_IRUSR
|S_IWUSR
, seccomp
),
2333 REG("mounts", S_IRUGO
, mounts
),
2335 REG("clear_refs", S_IWUSR
, clear_refs
),
2336 REG("smaps", S_IRUGO
, smaps
),
2338 #ifdef CONFIG_SECURITY
2339 DIR("attr", S_IRUGO
|S_IXUGO
, attr_dir
),
2341 #ifdef CONFIG_KALLSYMS
2342 INF("wchan", S_IRUGO
, pid_wchan
),
2344 #ifdef CONFIG_SCHEDSTATS
2345 INF("schedstat", S_IRUGO
, pid_schedstat
),
2347 #ifdef CONFIG_CPUSETS
2348 REG("cpuset", S_IRUGO
, cpuset
),
2350 INF("oom_score", S_IRUGO
, oom_score
),
2351 REG("oom_adj", S_IRUGO
|S_IWUSR
, oom_adjust
),
2352 #ifdef CONFIG_AUDITSYSCALL
2353 REG("loginuid", S_IWUSR
|S_IRUGO
, loginuid
),
2355 #ifdef CONFIG_FAULT_INJECTION
2356 REG("make-it-fail", S_IRUGO
|S_IWUSR
, fault_inject
),
2360 static int proc_tid_base_readdir(struct file
* filp
,
2361 void * dirent
, filldir_t filldir
)
2363 return proc_pident_readdir(filp
,dirent
,filldir
,
2364 tid_base_stuff
,ARRAY_SIZE(tid_base_stuff
));
2367 static struct dentry
*proc_tid_base_lookup(struct inode
*dir
, struct dentry
*dentry
, struct nameidata
*nd
){
2368 return proc_pident_lookup(dir
, dentry
,
2369 tid_base_stuff
, ARRAY_SIZE(tid_base_stuff
));
2372 static const struct file_operations proc_tid_base_operations
= {
2373 .read
= generic_read_dir
,
2374 .readdir
= proc_tid_base_readdir
,
2377 static const struct inode_operations proc_tid_base_inode_operations
= {
2378 .lookup
= proc_tid_base_lookup
,
2379 .getattr
= pid_getattr
,
2380 .setattr
= proc_setattr
,
2383 static struct dentry
*proc_task_instantiate(struct inode
*dir
,
2384 struct dentry
*dentry
, struct task_struct
*task
, const void *ptr
)
2386 struct dentry
*error
= ERR_PTR(-ENOENT
);
2387 struct inode
*inode
;
2388 inode
= proc_pid_make_inode(dir
->i_sb
, task
);
2392 inode
->i_mode
= S_IFDIR
|S_IRUGO
|S_IXUGO
;
2393 inode
->i_op
= &proc_tid_base_inode_operations
;
2394 inode
->i_fop
= &proc_tid_base_operations
;
2395 inode
->i_flags
|=S_IMMUTABLE
;
2397 #ifdef CONFIG_SECURITY
2398 inode
->i_nlink
+= 1;
2401 dentry
->d_op
= &pid_dentry_operations
;
2403 d_add(dentry
, inode
);
2404 /* Close the race of the process dying before we return the dentry */
2405 if (pid_revalidate(dentry
, NULL
))
2411 static struct dentry
*proc_task_lookup(struct inode
*dir
, struct dentry
* dentry
, struct nameidata
*nd
)
2413 struct dentry
*result
= ERR_PTR(-ENOENT
);
2414 struct task_struct
*task
;
2415 struct task_struct
*leader
= get_proc_task(dir
);
2421 tid
= name_to_int(dentry
);
2426 task
= find_task_by_pid(tid
);
2428 get_task_struct(task
);
2432 if (leader
->tgid
!= task
->tgid
)
2435 result
= proc_task_instantiate(dir
, dentry
, task
, NULL
);
2437 put_task_struct(task
);
2439 put_task_struct(leader
);
2445 * Find the first tid of a thread group to return to user space.
2447 * Usually this is just the thread group leader, but if the users
2448 * buffer was too small or there was a seek into the middle of the
2449 * directory we have more work todo.
2451 * In the case of a short read we start with find_task_by_pid.
2453 * In the case of a seek we start with the leader and walk nr
2456 static struct task_struct
*first_tid(struct task_struct
*leader
,
2459 struct task_struct
*pos
;
2462 /* Attempt to start with the pid of a thread */
2463 if (tid
&& (nr
> 0)) {
2464 pos
= find_task_by_pid(tid
);
2465 if (pos
&& (pos
->group_leader
== leader
))
2469 /* If nr exceeds the number of threads there is nothing todo */
2471 if (nr
&& nr
>= get_nr_threads(leader
))
2474 /* If we haven't found our starting place yet start
2475 * with the leader and walk nr threads forward.
2477 for (pos
= leader
; nr
> 0; --nr
) {
2478 pos
= next_thread(pos
);
2479 if (pos
== leader
) {
2485 get_task_struct(pos
);
2492 * Find the next thread in the thread list.
2493 * Return NULL if there is an error or no next thread.
2495 * The reference to the input task_struct is released.
2497 static struct task_struct
*next_tid(struct task_struct
*start
)
2499 struct task_struct
*pos
= NULL
;
2501 if (pid_alive(start
)) {
2502 pos
= next_thread(start
);
2503 if (thread_group_leader(pos
))
2506 get_task_struct(pos
);
2509 put_task_struct(start
);
2513 static int proc_task_fill_cache(struct file
*filp
, void *dirent
, filldir_t filldir
,
2514 struct task_struct
*task
, int tid
)
2516 char name
[PROC_NUMBUF
];
2517 int len
= snprintf(name
, sizeof(name
), "%d", tid
);
2518 return proc_fill_cache(filp
, dirent
, filldir
, name
, len
,
2519 proc_task_instantiate
, task
, NULL
);
2522 /* for the /proc/TGID/task/ directories */
2523 static int proc_task_readdir(struct file
* filp
, void * dirent
, filldir_t filldir
)
2525 struct dentry
*dentry
= filp
->f_path
.dentry
;
2526 struct inode
*inode
= dentry
->d_inode
;
2527 struct task_struct
*leader
= NULL
;
2528 struct task_struct
*task
;
2529 int retval
= -ENOENT
;
2532 unsigned long pos
= filp
->f_pos
; /* avoiding "long long" filp->f_pos */
2534 task
= get_proc_task(inode
);
2538 if (pid_alive(task
)) {
2539 leader
= task
->group_leader
;
2540 get_task_struct(leader
);
2543 put_task_struct(task
);
2551 if (filldir(dirent
, ".", 1, pos
, ino
, DT_DIR
) < 0)
2556 ino
= parent_ino(dentry
);
2557 if (filldir(dirent
, "..", 2, pos
, ino
, DT_DIR
) < 0)
2563 /* f_version caches the tgid value that the last readdir call couldn't
2564 * return. lseek aka telldir automagically resets f_version to 0.
2566 tid
= filp
->f_version
;
2567 filp
->f_version
= 0;
2568 for (task
= first_tid(leader
, tid
, pos
- 2);
2570 task
= next_tid(task
), pos
++) {
2572 if (proc_task_fill_cache(filp
, dirent
, filldir
, task
, tid
) < 0) {
2573 /* returning this tgid failed, save it as the first
2574 * pid for the next readir call */
2575 filp
->f_version
= tid
;
2576 put_task_struct(task
);
2582 put_task_struct(leader
);
2587 static int proc_task_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
2589 struct inode
*inode
= dentry
->d_inode
;
2590 struct task_struct
*p
= get_proc_task(inode
);
2591 generic_fillattr(inode
, stat
);
2595 stat
->nlink
+= get_nr_threads(p
);
2603 static const struct inode_operations proc_task_inode_operations
= {
2604 .lookup
= proc_task_lookup
,
2605 .getattr
= proc_task_getattr
,
2606 .setattr
= proc_setattr
,
2609 static const struct file_operations proc_task_operations
= {
2610 .read
= generic_read_dir
,
2611 .readdir
= proc_task_readdir
,